Azure II: Networking & AKS: glossary
104 terms, in plain words, with where each one is taught.
- 5-tuple hash (five-tuple)
- How a load balancer picks a backend: a hash of source IP, source port, destination IP, destination port and protocol, so one connection always lands on the same backend. Lesson 30.16
- AGIC (Application Gateway Ingress Controller, App Gateway for Containers)
- Application Gateway Ingress Controller: turns Kubernetes Ingress objects into App Gateway configuration. App Gateway for Containers is its successor. Lesson 30.16
- AKS tier (Free tier, Standard tier, Premium tier)
- The AKS pricing tier: Free (no SLA, for dev), Standard (uptime SLA, more API capacity, the production default), Premium (adds long-term support versions). Lesson 30.20
- allowForwardedTraffic (--allow-forwarded-traffic, forwarded traffic)
- A peering setting that accepts packets which did not start in the remote VNet but were forwarded through it, for example by the hub firewall. Lesson 30.14
- anycast
- One IP address announced from many locations at once, so each user reaches the nearest one. Lesson 30.16
- App Gateway listener (HTTP settings, request routing rule, listener (App Gateway))
- An App Gateway's chain: the listener (IP, port, host, certificate) -> a routing rule -> a backend pool plus HTTP settings (port, protocol, timeout) and a probe. Lesson 30.16
- Application Gateway (App Gateway, AGW, az network application-gateway)
- Azure's managed L7 reverse proxy: it reads each HTTP request's host and path, terminates TLS, and can run a WAF. Lesson 30.16
- ASG (application security group, application security groups)
- Application security group: a named group of NICs that NSG rules can use instead of IP ranges ("from asg-web to asg-db on 5432"). Not for pods. Lesson 30.3
- asymmetric routing
- Traffic going out one way and the replies coming back another. A stateful firewall that sees only half the connection drops it. Lesson 30.7
- auto-upgrade channel (planned maintenance, maintenance window)
- An AKS setting (patch, stable, node-image) that lets AKS upgrade on its own, inside a planned maintenance window you choose. Lesson 30.20
- autoscaler profile (autoScalerProfile, scale-down-unneeded-time, scan-interval)
- Cluster-wide settings for the AKS cluster autoscaler: how often it checks, how long a node must be unneeded before removal, which expander it uses. Lesson 30.25
- az aks (az aks show, az aks nodepool list, az aks nodepool add)
- The az command group for AKS: show a cluster, and list, add, update, scale and delete its node pools (
--cluster-name,-n <pool>). Lesson 30.20 - az aks create (--vnet-subnet-id, --generate-ssh-keys, --node-count)
- Creates an AKS cluster.
--network-plugin/--network-plugin-modepick the network mode,--vnet-subnet-idthe node subnet,--node-counthow many nodes. Lesson 30.22 - az aks get-credentials (get-credentials, --admin)
- Downloads a kubeconfig for an AKS cluster and merges it into ~/.kube/config. With
--admin, the local cluster-admin certificate that bypasses Entra ID. Lesson 30.31 - az aks rotate-certs (rotate-certs)
- Replaces every certificate of an AKS cluster, invalidating all certificate-based kubeconfigs ever handed out. Restarts the control plane and nodes. Lesson 30.31
- az aks stop (az aks start)
- Stops an AKS cluster's nodes and control plane so you stop paying for the VMs;
az aks startbrings it back with new node VMs. For dev clusters at night, never to restart production. Lesson 30.20 - az aks upgrade (az aks get-upgrades, --control-plane-only, az aks nodepool upgrade)
- Upgrading AKS:
get-upgradeslists allowed versions,az aks upgrade -k <version>moves the control plane (alone with--control-plane-only),az aks nodepool upgrademoves one pool. - az network vnet (az network vnet show, az network vnet subnet create, az network vnet subnet show)
- The az commands for VNets and their subnets: show a VNet, create/show/update a subnet (
--address-prefixes,--nsg,--route-table). Lesson 30.1 - Azure Bastion (AzureBastionSubnet)
- Azure's managed jump host: you reach VMs through it over SSH or RDP instead of giving them public IPs. Lesson 30.1
- Azure CNI Overlay (--network-plugin-mode overlay, overlay mode)
- The default AKS network mode: pods get addresses from a private pod CIDR (a /24 per node) and are NAT-ed to the node IP when they leave. One VNet address per node. Lesson 30.22
- Azure Firewall (AzureFirewallSubnet)
- Azure's managed firewall. It lives in its own subnet, which must be called AzureFirewallSubnet, and can allow traffic by FQDN as well as by IP. Lesson 30.1
- Azure Load Balancer (Standard Load Balancer, az network lb)
- Azure's L4 load balancer: it forwards TCP/UDP connections to a pool of backends by address and port. Every AKS LoadBalancer Service uses one. Lesson 30.16
- Azure RBAC for Kubernetes (--enable-azure-rbac, Azure Kubernetes Service RBAC Reader, Azure Kubernetes Service RBAC Writer)
- An AKS mode where the API server asks Azure whether a token may do something, using role assignments scoped to the cluster or one namespace (RBAC Reader, Writer, Admin, Cluster Admin). Lesson 30.31
- Azure reserved addresses (five reserved addresses, 2^(32-n) - 5)
- Azure keeps 5 addresses of every subnet: the network address, .1 (the gateway), .2 and .3 (Azure DNS) and the last one. Usable = 2^(32-prefix) - 5, and the first address you get is .4. Lesson 30.1
- backend health (show-backend-health, healthProbeLog)
- App Gateway's view of each backend server: Healthy or Unhealthy, with the probe's reason. The first thing to read on any 502. Lesson 30.16
- backend pool (backend address pool)
- The set of servers (IPs or FQDNs, or the VMs behind a load balancer) traffic is shared between. Lesson 30.16
- Cluster User Role (Azure Kubernetes Service Cluster User Role, listClusterUserCredential)
- The Azure role that lets you download an AKS kubeconfig through ARM. It gives no rights inside the cluster. Lesson 30.31
- CriticalAddonsOnly (CriticalAddonsOnly=true:NoSchedule, only_critical_addons_enabled)
- The taint you put on a system pool so only cluster add-ons run there and your apps stay off it. Lesson 30.20
- custom DNS server (DNS forwarder, conditional forwarder)
- A DNS server you run (on-prem or a VM) instead of Azure DNS. It must forward privatelink.* questions to 168.63.129.16 or private endpoints resolve publicly. Lesson 30.10
- data exfiltration (exfiltration)
- Someone copying your data out to a place they control. Lesson 30.10
- default NSG rules (AllowVnetInBound, AllowAzureLoadBalancerInBound, DenyAllInBound)
- The six rules every NSG has at priority 65000-65500: allow the whole VNet in and out, allow load balancer probes, allow internet out, deny the rest. An empty NSG still lets the VNet in. Lesson 30.3
- disable local accounts (--disable-local-accounts, disableLocalAccounts)
- An AKS setting that stops new admin certificates being issued, so everyone must log in through Entra ID. Existing ones need a certificate rotation. Lesson 30.31
- DNS resolver (Private DNS Resolver)
- A server that answers DNS questions for other machines; in a hub it answers for all the spokes. Lesson 30.14
- DNS zone group (zone group, dns-zone-group)
- An attachment on a private endpoint that creates and deletes its A record in the private DNS zone automatically, so the record never goes stale. Lesson 30.10
- dns-service-ip (--dns-service-ip, --service-cidr)
- The ClusterIP CoreDNS gets (default 10.0.0.10), inside the service CIDR (default 10.0.0.0/16). Neither may overlap the VNet. Lesson 30.22
- effective routes (show-effective-route-table)
- The merged result of system, BGP and user routes that really apply to a NIC, with where each came from.
az network nic show-effective-route-table. Lesson 30.7 - eviction policy (spot) (--eviction-policy, --spot-max-price, spot max price)
- Spot pool settings:
--eviction-policy Deletedeletes evicted VMs;--spot-max-price -1pays up to the on-demand price, so you are evicted only for capacity. Lesson 30.25 - exec plugin (exec credential plugin)
- A program named in a kubeconfig that kubectl runs to fetch a fresh token every time, instead of storing a credential in the file. Lesson 30.31
- expander (priority expander, least-waste)
- How the cluster autoscaler picks which pool to grow when several fit: random, least-waste, or priority (a ranking such as "spot first, then regular"). Lesson 30.25
- ExpressRoute (GatewaySubnet, VPN gateway)
- A private line between a company's own network and Azure. Its gateway (and the VPN gateway, the encrypted-tunnel option) lives in a subnet called GatewaySubnet. Lesson 30.1
- ForbiddenByConnection
- Key Vault's error when the network path is refused (public access off, caller not on a private endpoint) - a network problem, not RBAC.
- forced tunnelling (forced tunneling)
- Sending all traffic, even internet-bound, back through on-prem or a firewall instead of straight out. Lesson 30.7
- FQDN rule (FQDN rules, application rule)
- A firewall rule that allows traffic by host name (mcr.microsoft.com) rather than by IP address. Lesson 30.7
- Front Door (Azure Front Door)
- Azure's global L7 edge in front of your apps: a CDN, a WAF and anycast in one. Lesson 30.16
- gateway transit (--allow-gateway-transit, --use-remote-gateways)
- Letting spokes use the hub's VPN or ExpressRoute gateway through their peering, so on-prem reaches every spoke through one gateway. Lesson 30.14
- GPU
- Graphics processing unit: a chip for heavy parallel maths (machine learning, video). GPU VM sizes are expensive and get their own node pool.
- group id (private endpoint) (--group-id, sub-resource)
- Which part of a resource a private endpoint is for: blob, file, vault, registry... One endpoint per sub-resource. Lesson 30.10
- health probe (health probes, 168.63.129.16)
- A load balancer regularly sending a test connection or request to each backend; backends that fail get no traffic. Azure's probes come from 168.63.129.16. Lesson 30.3
- hub VNet (spoke VNet, hub, spoke)
- In hub and spoke, the hub is the central VNet with shared services (firewall, gateways, DNS); spokes are the teams' VNets, each peered only with the hub. Lesson 30.1
- inbound (outbound, direction (NSG))
- Inbound is traffic coming into a subnet or NIC, outbound is traffic leaving it. NSG rules are evaluated per direction. Lesson 30.3
- InsufficientSubnetSize
- AKS's error when a scale-up, upgrade or create would need more pre-allocated IPs than the subnet has. The message contains the arithmetic. Lesson 30.22
- konnectivity
- The tunnel between AKS's control plane (in Microsoft's subscription) and your nodes. Lesson 30.20
- kubelogin (kubelogin convert-kubeconfig)
- Microsoft's exec plugin that gets Entra ID tokens for kubectl.
convert-kubeconfig -l azureclireuses your az login;-l workloadidentityis for automation. Lesson 30.31 - kubenet
- The older, simpler AKS network plugin: pods use a private range routed with a route table. Being retired; do not start new clusters on it. Lesson 30.22
- LB
- Short for load balancer; on Azure usually the Azure Load Balancer. Lesson 30.16
- managed control plane (AKS control plane)
- On AKS, the API server, etcd, scheduler and controller manager run in Microsoft's subscription. You see one endpoint, never the machines. Lesson 30.20
- Microsoft backbone (backbone)
- Microsoft's own network between its data centres. Private endpoint traffic stays on it instead of crossing the internet. Lesson 30.10
- NAT Gateway (NAT gateway)
- A dedicated Azure resource for outbound SNAT, attached to a subnet, with far more ports than a load balancer's outbound rule. Lesson 30.16
- NetcfgSubnetRangesOverlap (NetcfgSubnetRangeOutsideVnet)
- Azure's errors for a subnet whose range overlaps another subnet, or lies outside the VNet's address space. Lesson 30.1
- Network Watcher (IP flow verify, test-ip-flow)
- Azure's network diagnostics service. IP flow verify answers "would this packet be allowed, and by which rule?" for a VM NIC. Lesson 30.3
- next hop type (VirtualAppliance, VirtualNetworkGateway, NextHopType)
- Where an Azure route sends packets: VirtualAppliance (an IP such as a firewall), VirtualNetworkGateway (VPN/ExpressRoute), VnetLocal, Internet, or None (drop). Lesson 30.7
- node image upgrade (--node-image-only, node image)
- Replacing a pool's nodes with a newer OS image (security patches) without changing the Kubernetes version. Lesson 30.20
- node resource group lockdown (--nrg-lockdown-restriction-level)
- An AKS setting that puts a deny assignment on the node resource group so nobody can change AKS-managed resources by hand. Lesson 30.20
- node subnet mode (Azure CNI node subnet, legacy Azure CNI, node-subnet)
- The older Azure CNI mode: each node pre-allocates maxPods + 1 VNet addresses, so pods are directly reachable in the VNet but subnets fill up fast. Lesson 30.22
- NotTriggerScaleUp (TriggeredScaleUp, FailedScaleUp)
- The cluster autoscaler's events on Pending pods: it scaled up, decided no pool could help (and why), or Azure refused the new node. Lesson 30.25
- NSG flow logs (VNet flow logs, flow log)
- Records of every allowed and denied connection through an NSG (now per VNet), written to a storage account. Lesson 30.3
- NSG priority (priority (NSG))
- The number on each NSG rule (100-4096 for yours). Rules are checked lowest number first and the first match wins. Two rules in one direction cannot share a number. Lesson 30.3
- NTP
- Network Time Protocol: how machines keep their clocks right. Nodes behind a firewall must be allowed to reach a time server. Lesson 30.7
- NVA (network virtual appliance)
- Network virtual appliance: a VM running firewall or router software that traffic is sent through, as an alternative to Azure Firewall. Lesson 30.1
- on-demand (regular priority)
- A normal VM at full price that Azure never evicts, as opposed to spot. Lesson 30.25
- outbound rule (aksOutboundRule, az network lb outbound-rule)
- A load balancer rule that carries backends' traffic out to the internet through its public IPs (SNAT). AKS uses one with a 30-minute idle timeout. Lesson 30.16
- OWASP
- A community project that catalogues common web attacks; WAF rule sets are built from its list. Lesson 30.16
- PaaS (platform as a service)
- Platform as a service: a service Azure runs for you, such as Key Vault or a storage account, as opposed to a VM you run yourself. Lesson 30.10
- PE
- Short for private endpoint, often used in names (pe-blob, snet-pe). Lesson 30.10
- peering state (PeeringState, Initiated, Connected)
- Initiated means only one side of a peering exists and nothing flows; Connected means both sides exist. Lesson 30.14
- pod subnet mode (--pod-subnet-id, dynamic IP allocation)
- An Azure CNI mode where pods take VNet addresses from a separate pod subnet, handed out in blocks of 16 as needed. Lesson 30.22
- pre-allocated IPs (pre-allocated)
- Addresses reserved the moment a node is created, whether or not pods use them. On node-subnet CNI each node reserves maxPods + 1. Lesson 30.22
- Preempt event (scheduled event, node problem detector)
- Azure's ~30-second warning before evicting a spot VM. AKS's node problem detector turns it into a PreemptScheduled event on the node. Lesson 30.25
- private cluster (--enable-private-cluster, az aks command invoke)
- An AKS cluster whose API server sits behind a private endpoint. kubectl works only from networks that reach it, or through
az aks command invoke. Lesson 30.31 - private DNS zone (privatelink zone, privatelink.blob.core.windows.net, privatelink.vaultcore.azure.net)
- A DNS zone only the VNets linked to it can see. A privatelink.* zone holds the A record that makes a service's normal name resolve to its private endpoint. Lesson 30.10
- Private Link (privatelink)
- The Azure feature behind private endpoints: reaching one PaaS resource over a private IP in your VNet, on Microsoft's own network. Lesson 30.10
- route table (Azure) (route table, az network route-table, az network route-table route create)
- An Azure resource holding your own routes; attached to a subnet, it adds or overrides routes for traffic leaving that subnet. Lesson 30.7
- scalesetpriority (kubernetes.azure.com/scalesetpriority, kubernetes.azure.com/scalesetpriority=spot:NoSchedule)
- The label and taint AKS puts on spot nodes by itself, so nothing runs there unless it tolerates the taint. Lesson 30.25
- security rule (NSG rule, az network nsg rule create, az network nsg rule list)
- One allow or deny line of an NSG: source, source port, destination, destination port, protocol, direction and priority. Lesson 30.3
- SecurityRuleConflict
- Azure's error when you create an NSG rule with a priority another rule in the same direction already has. Lesson 30.3
- service tag (service tags, VirtualNetwork, AzureLoadBalancer)
- A name in an NSG rule that stands for a set of IP ranges Microsoft keeps up to date, such as Internet, VirtualNetwork, AzureLoadBalancer or GatewayManager. Lesson 30.3
- split-brain (split brain)
- Two copies of something that should be one, disagreeing - for example a private DNS zone created separately in every subscription. Lesson 30.10
- spot VM (spot, spot pool, spot node pool)
- A VM on Azure's spare capacity at up to ~90% off, which Azure can evict at any time with about 30 seconds' notice. For interruptible work only. Lesson 30.25
- SQL (SQL injection)
- The language relational databases are queried with. SQL injection is an attack that sneaks SQL into a request field. Lesson 30.16
- stateful filtering
- A filter that remembers connections, so replies to an allowed connection are allowed automatically. NSGs are stateful: you never write the reply rule. Lesson 30.3
- subnet delegation (delegations)
- Handing a whole subnet to one Azure service (for example a managed database) so that only that service may put things in it. Lesson 30.1
- SubnetIsFull
- Azure's error when a subnet has no free address left for a new NIC or private endpoint.
- System node pool (User pool, mode (node pool), --mode)
- A node pool in System mode runs the cluster's own pods (CoreDNS and friends) and at least one is required; User pools run your workloads. Lesson 30.20
- system route (system routes, VnetLocal)
- A route Azure creates in every subnet by itself: the VNet's own range (VnetLocal), peered ranges, 0.0.0.0/0 to the internet, and private ranges dropped. Lesson 30.7
- TCP reset (load balancer) (enableTcpReset, TCP reset)
- A load balancer setting: when an idle flow is forgotten, send RST to both ends so clients fail fast and reconnect instead of hanging. Lesson 30.16
- transitive (not transitive, transit)
- If A reaches B and B reaches C, then A reaches C. Peering is not transitive: spoke A and spoke B both peered with the hub still cannot talk without a router in the hub. Lesson 30.1
- UDR (user-defined route, user-defined routes)
- User-defined route: one route you add to a route table. On a tie in prefix length it beats BGP and system routes. Lesson 30.7
- userDefinedRouting (--outbound-type userDefinedRouting, outbound type)
- An AKS create option: the cluster makes no outbound public IP and relies on your route table to send egress through a firewall. Lesson 30.7
- Virtual WAN (vWAN, AVNM, Virtual Network Manager)
- Microsoft's managed hub service, with routing between spokes built in. AVNM (Virtual Network Manager) creates connectivity and pushes security rules centrally. Lesson 30.14
- VM size names (Standard_D8ds_v5, Standard_D4ds_v5)
- How Azure VM sizes read: Standard_D8ds_v5 = D family (general purpose), 8 vCPUs, d = local disk, s = premium storage, v5 = generation. Lesson 30.20
- VNet link (virtual network link, az network private-dns link vnet create, --registration-enabled)
- The object that lets one VNet see a private DNS zone. No link, and the name resolves to the public IP. Lesson 30.10
- VNet peering (peering, peered, az network vnet peering)
- A direct connection between two VNets so their addresses can reach each other. Made of two objects, one on each side, and not transitive. Lesson 30.1
- WAF (web application firewall, WAF_v2)
- Web application firewall: rules that inspect HTTP requests and block common attacks such as SQL injection. On Azure it runs on Application Gateway or Front Door. Lesson 30.3
- WAN (wide area network)
- Wide area network: a network that connects sites far apart, such as a company's offices and its cloud. Lesson 30.14